Probabilistic safety assessment of multi-functional flood defences

(1)

Juan Pablo Aguilar López

Probabilistic Safety Assessment of

Multi-functional Flood Defences

(2)

Probabilistic Safety Assessment of

Multi-functional Flood Defences

(3)

Promotion committee:

Prof. dr. ir. G.P.M.R. Dewulf University of Twente, Chairman/Secretary Prof. dr. S.J.M.H. Hulscher University of Twente, Promotor

Dr. J.J. Warmink University of Twente, Co-promotor

Dr. R.M.J. Schielen University of Twente, Co-promotor Prof.dr.ir. M. Kok Technical University of Delft Prof. ir. A.C.W.M. Vrouwenvelder Technical University of Delft Prof.dr. M. van der Meijde University of Twente Prof.dr. R.W.M.R.J. Ranasinghe University of Twente Other members:

dr. J.B. Sellmeijer

This research is supported by the Dutch Technology Foundation STW, which is part of the Netherlands Organisation for Scientific Research (NWO), and which is partly funded by the Ministry of Economic Affairs (Project number: P10-28/12178).

This research is part of the program on Integral and Sustainable Design of Multifunctional Flood Defences.

Cover design: Juan Pablo Aguilar López

Copyright © 2016 by Juan Pablo Aguilar López, Enschede, the Netherlands. All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, without the written permission of the author.

Printed by Gildeprint

DOI: 10.3990/1.9789036542586 ISBN 978-90-365-4258-6

(4)

PROBABILISTIC SAFETY ASSESSMENT

OF MULTI-FUNCTIONAL FLOOD DEFENCES

DISSERTATION

to obtain

the degree of doctor at the University of Twente, on the authority of the rector magnificus

Prof.dr. T.T.M. Palstra,

on account of the decision of the graduation committee, to be publicly defended

on Friday the 9th_{of December 2016 at 16:45hrs}

by

Juan Pablo Aguilar López

born on the 9th_{of December 1981}

(5)

This thesis was approved by:

Prof. dr. S.J.M.H. Hulscher Promotor

Dr. J.J. Warmink Co-promotor

(6)

‘By three methods we may learn wisdom: First, by reflection, which is the

noblest; Second, by imitation, which is the easiest; and third by experience,

which is the bitterest.’

Confucius (551 BC - 479 BC)

(7)

(8)

Preface

The children’s book called “Hans Brinker or the silver skates” was written by the American author Mary Mapes Dodge and first time published in 1865. The main plot developed around the story of a poor and honourable young boy named Hans Brinker and his younger sister Gretel who eagerly wanted to participate in December’s great ice skating race on the frozen canals. The book was originally subtitled as a “story of life in Holland” and aimed to depict the Dutch life on the 1800’s to the young American readers. The author dedicates chapter number two, to try to describe in detail, the atypical and yet fascinating landscapes and daily life of what she refers to as

Contrary-land. A place which has a vast portion of the land, laying below the level of the sea.

Where dikes, ditches and canals are everywhere to be seen. Where often the keels of floating ships are higher than the roofs of the dwellings. And she questions: “Which is Holland? The shores or the water?”. Since those early times, it was already acknowledged that living in the Netherlands meant to have the blue colour always present in the lower parts of the landscape as well. In the same manner, this same landscape could be suddenly split in two by a green and brownish stripe where people tend to go whenever the waters got angry.

(9)

“Angry” was the expression used by the father of the main character in “The Hero of Haarlem”, a child’s story taught as lesson 62 in school during the English reading class in the same Novel.

This inner story described how a little boy saved the town of Haarlem by putting his finger inside a leak of a dike in order to avoid an imminent catastrophe. He spent the whole night in the freezing cold retaining water flow until he was found in the next morning by a clergyman who was walking back home on a trail located over the dike. In that way, the little boy saved the town and the story inspired the young readers in the novel.

Hero of Haarlem. Illustration extracted from “Hans Brinker or the Silver Skates, Dodge (1876)”

The story tells, that Mr. Raff Brinker had been employed for several years for the maintenance of the dikes which kept the city safe during the angry water times. After identifying a weak leakage spot near the Veermyk sluice, the midst of a severe storm impaired the worker’s vision making him fall from a scaffolding. In this sense, the story also reflects the importance of the dikes to the Dutch community and how since the early days of Dutch history flood risk has represented a major threat to life. Nevertheless, the early Dutch way of living did not succumb to fear but on the contrary it has defied the “angry waters” by using the dikes as habitable spaces and/or access pathways as well. Both traits are also described by Dodge, who also stated in her book

(10)

are high and wide, and the tops of some of them are covered with buildings and trees. They have even fine public roads on them, from which horses may look down upon wayside cottages”.

After all previous highlights of all sort of flood defence technicalities, is natural that the young reader’s mind will be intrigued by one main question:

“ If dikes had structures of considerable size such as buildings and roads embedded, what are the chances

that the whole flood defence reliability can be changed by one kid’s finger ? “

The present study aims to elucidate a very small and yet significant part of this complex question which directly relates to the influence of structural embedments, in the probabilistic design and assessment of a multi-functional flood defence. Note that the fundamental support for the idea of expressing a system performance in terms of a probabilistic measure comes from the acknowledgement of its uncertain nature. The book of Hans Brinker on its own becomes a large source of uncertainty if taken into account that the author had never visited the Netherlands before publishing the novel.

(11)

(12)

Summary

The state of the art related to flood defences design and assessment show that the most common adaptation measures for tackling the uncertainties regarding climate change and demographic explosion, often involve an increase in size of these water-retaining structures. The shortage of habitable areas in the deltaic zones and the required increase in dimensions gave birth to the multifunctional flood defence concept. The main idea behind the multifunctional flood defences is that the extra space that results from almost inevitable increase in dimensions can be exploited by including additional non-water retaining functions such as commercial, recreational, ecological and habitational space. The inclusion of these new functions will also allow these defences to be financially feasible as their marginal cost reduces due to the expected benefits from the additional function allocation. The addition of functions will require structural embedments for connecting them to the transport, sewage, water supply, and electricity networks. Flood defences are exposed to deterioration processes known as “failure mechanisms” which may be triggered during flood events. The embedment of hard structures derived from the connecting requirements to the different infrastructure networks will have an impact on these processes and their frequency of occurrence. In the last decades, due to the advance in computing capability, the structural design discipline has migrated towards the implementation of probabilistic safety assessments for existing structures and for new designs. Moreover, probabilistic target values are now being included in the latest flood risk policies and normative codes. Yet, most of these values are defined based on the results of probabilistic assessments performed for flood defences which have no other function besides the water retaining one. The inclusion of non-water retaining structures will have an impact in most of the failure mechanisms and consequently their safety estimated values.

In the particular case of the Netherlands, the VNK project which is the largest national flood risk assessment study, concluded that despite the large number of potential failure mechanisms only a few of them account for most of the estimated failure probability given the actual state of the flood defence system. Especially, the ones which were described by an erosive process such as piping and wave overtopping grass cover failure. This does not mean that only these two failure mechanisms are important to assess or that they are the most frequently registered cases. It also means that the associated uncertainty to these failure mechanisms may be larger when compared to the uncertainty associated with other failure mechanisms.

(17)

defence reliability. Consequently, they should be prioritized for a deeper study over the ones that influence non-erosion based failure mechanisms. Yet, it is acknowledged that all failure mechanisms are important. As a result, the objective of the present thesis is “

To determine and quantify the effects in the probabilistic design and safety assessment of erosion based failure mechanisms of MFFD’s, derived from the foundation and the embedment of hard structures ”.

The research was divided in two main components. The first part aimed to study the effects derived from the design and assessment choice of reducing the associated uncertainty of the materials involved in the failure mechanisms modelling. The second part aimed to study the effects derived from geometrical design choices of the embedded structure such as size and location. In order to do that, one case studies was used for the first part and two more for the second part.

In Chapter 2, the first case study correspond to the assessment of the effects on the probabilistic safety assessment from reducing the material uncertainty by including the “potential” correlation effect between the representative grain size d70 and the

foundation aquifer hydraulic conductivity variable K. This is done for the piping erosion failure mechanism assessed via the revised Sellmeijer limit state equation. The results showed that correlation between d70 and K has a significant effect on the failure

probability of structures for the case of piping erosion. This is because it increases the probability of having small d70 grains in combination with large hydraulic conductivity

values which analogously reduces the probability high conductivity values in combination with small d70 grain diameters. These effects are even more important for

multifunctional flood defences as they can help to reduce the minimum required seepage length for achieving its safety target values.

The second case study is presented in Chapter 3 and corresponds to the assessment of the effects on the probabilistic safety assessment for piping erosion from embedding a sewer pipe under a flood defence. This assessment is done by implementing a finite element model in combination with emulation techniques which allow the modification of the sewer pipe characteristics for assessing the effects of size and location. The results showed that for the studied case, the embedment of sewer pipes inside the aquifer foundation of multifunctional flood defences will always represent an increase in safety against piping erosion. The degree of improvement of the safety is conditioned to the location of the pipe, the size of the pipe, the aquifer depth, the aquifer confinement and the aquifer equivalent hydraulic conductivity.

The third and final case study is presented in Chapter 4. This case study consists of the assessment of a flood defence exposed to wave overtopping which threatens its safety due to the erosion of the grass cover. The effects on the failure probability due to these failure mechanisms are assessed by including the change in the overtopped wave’s

(18)

techniques. In addition, the grass cover quality spatial variability is also included in the analysis which allows to determine which zones are more prone to failure for dikes with and without a road. The results showed that the presence of asphalt roads will reduce the safety against grass cover erosion due to wave overtopping. This reduction is explained by the change in superficial material roughness and the presence of superficial profile irregularities which increase the likelihood of localized failures. In addition, it is also observed how the grass quality spatial variability and the stochastic nature of the localized resistance variables are more important indicators of potential failure for dikes with roads on top.

In Chapter 5, the applicability of the methods and the implications of the assumptions were further discussed. In terms of implications it is shown that the correlation omission may result in wider structures. In addition, the implementation of the new Sellmeijer limit state equation may not be suitable in its actual state for assessing defences with sewer pipes underneath. In relation with the grass erosion due to overtopping, the erodability values of grass should be represented as functions of the critical shear stress for different grass qualities instead of constant values for each grass quality. In order to compare the importance of each choice in the total safety of the flood defence, the change in the reliability index was estimated for each choice.

Finally in Chapter 6, the main conclusions are compiled and further recommendations are listed. The main outcome of this thesis is that the inclusion of hard structures in the flood defence have significant effects in their reliability which is not negligible. These effects are derived from design uncertainties associated to the materials and the dimensioning of the multifunctional flood defences. Furthermore, it was observed that these effects may be either positive or negative in percentages of as much as 20% for the studied cases, with respect to the case of an identical flood defence which doesn’t have an embedded structure. Therefore, it is recommended to include this effects in the future safety assessments of multifunctional flood defences and to update the actual design tools so that the additional embedded structures may be included.

(19)

Samenvatting

Het tekort aan bewoonbare gebieden in delta’s en de vereiste toename in afmetingen van waterkeringen gezien klimaatverandering, heeft geleid tot het ontstaan van het multifunctionele waterkering (MFWK) concept. Het belangrijkste idee achter dit concept is dat de extra ruimte, die gecreëerd wordt als gevolg van de quasi onvermijdelijke toename in afmetingen van toekomstige waterkeringen, geëxploiteerd kan worden met extra niet-waterkerende functies zoals commerciële, recreatieve, ecologische en woonruimte. De opname van deze extra functies maakt grotere waterkeringen financieel haalbaar aangezien marginale kosten verminderen door de verwachte voordelen van deze extra functies. Deze extra functies vereisen echter ook structurele inbeddingen en funderingen om deze ruimte aan te sluiten op het vervoer, riolering, waterleiding, en elektriciteitsnetwerken. Deze extra verankering zal de veiligheid van de multifunctionele waterkering veranderen ten opzichte van conventionele waterkeringen van dezelfde afmetingen.

Als gevolg van de toegenomen rekencapaciteit, is structureel ontwerp als discipline in de laatste decennia gemigreerd naar de uitvoering van probabilistische veiligheidsbeoordelingen van structuren. Bovendien worden probabilistische streefwaarden nu ook opgenomen in beleid rond overstromingsrisico en in normatieve codes gezien de nood aan officiële risico-gebaseerde veiligheidsnormen. Deze evaluaties bestaan uit het inschatten van de faalkans van structuren of systemen opgebouwd uit structuren als gevolg van hun belangrijkste faalprocessen. Deze processen zijn bekend als "faalmechanismen" die geactiveerd kunnen worden tijdens overstromingen. Het opnemen van structuren naast de waterkeringen zelf, zal een impact hebben op het optreden van deze faalmechanismen en bijgevolg op de geschatte veiligheidswaarden. Voor het specifieke geval van Nederland, hebben verschillende studies rond overstromingsrisico geconcludeerd dat ondanks het grote aantal potentiële faalmechanismen slechts een paar mechanismen instaan voor het grootste deel van de geschatte faalkans gezien de werkelijke toestand van de waterkering. Het gaat dan vooral over faalmechanismen die toegeschreven worden aan een erosief proces, zoals “piping” en falen van grasmatten door golfoverslag. De scope van de huidige studie is bijgevolg verfijnd, gebaseerd op de veronderstelling dat ontwerpkeuzes die de erosie gerelateerde faalmechanismen direct beïnvloeden, de grootste gevolgen zouden hebben op de totale betrouwbaarheid van de multifunctionele waterkering. Daarom moeten deze ontwerpkeuzes geprioriteerd worden voor een diepere studie boven ontwerpkeuzes die niet-erosie gerelateerde faalmechanismen beïnvloeden. Toch wordt erkend dat alle faalmechanismen belangrijk zijn.

(20)

Als gevolg daarvan is het doel van dit proefschrift "Het determineren en kwantificeren van de

effecten van erosie-gerelateerde faalmechanismen van MFWK's, afkomstig van de fundering en inbedding van harde structuren, op het probabilistische ontwerp en de veiligheidsbeoordeling ".

Het onderzoek is opgesplitst in twee hoofdonderdelen. Het eerste deel bestudeert het effect van ontwerp en evaluatie keuzes gericht op het verminderen van de onzekerheid in materialen bij het modelleren van faalmechanismen. Het tweede deel bestudeert het effect van geometrisch ontwerp keuzes van de ingebedde structuren, zoals de grootte en de locatie. Hiervoor werden casus studies gebruikt; een voor het eerste deel en twee voor het tweede deel.

In hoofdstuk 2, wordt in de eerste casus het effect bestudeerd van verminderde materiaalonzekerheid op de probabilistische veiligheidsbeoordeling door inbegrip van het "potentiele" correlatie effect tussen korrelgrootte d70 en de fundering aquifer geleidbaarheid variabele K. Dit is gedaan voor het faalmechanisme gerelateerd aan pijperosie dat geëvalueerd wordt met de herziene Sellmeijer grenstoestand vergelijking. De resultaten tonen aan dat de correlatie tussen d70 en K een significant effect heeft op de faalkans van structuren bij pijperosie. Dit is omdat het de kans op kleine d70 korrels in combinatie met grote K waarden verhoogt en in analogie de kans op hoge K in combinatie met kleine diameters d70 korrel vermindert. Deze effecten zijn nog belangrijker voor multifunctionele waterkeringen aangezien ze de minimaal vereiste kwelweglengte voor het bereiken van de veiligheid streefwaarden kunnen helpen verminderen.

De tweede casus wordt gepresenteerd in hoofdstuk 3 en betreft de beoordeling van het effect van een onder de waterkering ingebedde rioolbuis op de probabilistische veiligheidsbeoordeling voor pijperosie. Deze beoordeling gebeurt op basis van een eindige elementen model in combinatie met emulatietechnieken die variatie van de rioolbuis kenmerken toelaten om de effecten van grootte en locatie te evalueren. De resultaten tonen aan dat voor de onderzochte casus, de inbedding van rioolbuizen in de aquifer fundering van multifunctionele waterkeringen altijd een toegenomen veiligheid tegen pijperosie vertegenwoordigt. De mate waarin de veiligheid toeneemt, wordt geconditioneerd door de locatie van de buis, de diameter van de buis, de aquifer diepte, de aquifer begrenzing en de aquifer equivalente geleidbaarheid.

De derde en laatste casus wordt gepresenteerd in hoofdstuk 4. In deze casus wordt een waterkering geëvalueerd die blootgesteld wordt aan golfoverslag waarbij de veiligheid wordt bedreigd als gevolg van de erosie van de grasmat. De effecten van deze faalmechanismen op de faalkans worden beoordeeld met inbegrip van de veranderende hydrodynamica van de overslaande golf op basis van computermodellen in combinatie met emulatie technieken. Bovendien wordt de ruimtelijke variatie in grasmat kwaliteit

(21)

meer kans op falen hebben en dit voor dijken met en zonder weg. De resultaten tonen aan dat de aanwezigheid van asfaltwegen de veiligheid tegen grasmat erosie door golfoverslag zal verminderen. Deze daling wordt verklaard door de verandering in ruwheid van het oppervlakte materiaal en de aanwezigheid van onregelmatigheden in het oppervlakte profiel waardoor de kans op falen gelokaliseerd verhoogt. Daarnaast wordt ook waargenomen dat de ruimtelijke variabiliteit in graskwaliteit en de stochastische aard van gelokaliseerde resistentie variabelen belangrijke indicatoren zijn voor mogelijk falen van dijken waarop wegen lopen.

In hoofdstuk 5 worden de toepasbaarheid van de methoden en de gevolgen van de aannames verder besproken. Wat betreft gevolgen van aannames, wordt aangetoond dat het weglaten van correlatie tussen korrelgrootte en geleidbaarheid kan leiden tot bredere structuren. Bovendien blijkt dat de nieuwe Sellmeijer grenstoestand vergelijking in zijn huidige vorm mogelijk niet geschikt is voor de beoordeling van waterkeringen waaronder rioolbuizen lopen. Wat betreft graserosie door golfoverslag, moet de erosiegevoeligheid van gras weergegeven worden als functie van de kritische schuifspanning voor verschillende graskwaliteit in plaats van te werken met constante waarden voor elke graskwaliteit. Om het belang van elke keuze te vergelijken in de totale waterkering veiligheid, werd de verandering in de betrouwbaarheidsindex geschat voor elke keuze.

Tot slot worden in hoofdstuk 6 de belangrijkste conclusies gecompileerd en verdere aanbevelingen opgenomen. Het belangrijkste resultaat van dit proefschrift is dat de opname van harde structuren in de waterkering een significant effect hebben op hun betrouwbaarheid en dit effect is niet te verwaarlozen. Dit significant effect stamt uit ontwerp onzekerheden rond de materialen en de dimensionering van de multifunctionele waterkeringen. Voorts werd opgemerkt dat deze effecten zowel positief als negatief kunnen zijn en dit tot in 20% van de onderzochte gevallen, in vergelijking met een identieke waterkering zonder toegevoegde structuur. Daarom verdient het aanbeveling om deze effecten op te nemen in de toekomstige veiligheidsbeoordeling van multifunctionele waterkeringen en om de gebruikte ontwerptools bij te werken, zodat de extra ingebedde structuren erin opgenomen kunnen worden.

(22)

1. Introduction

1.1 Multi-functional flood defences (MFFD’s)

Multifunctional flood defences (MFFD’s), are being conceived as an innovative and yet robust solution where flood defence components can be safely combined with non-water retaining components (Van Loon-Steensma and Vellinga, 2014). Sea level rise and climate change have proven to reduce the effectiveness of flood defence structural measures and consequently increase the risk of flooding (Klijn et al., 2015). The increase of dimensions is the most common adaptation measure for tackling the uncertainties that affect the safety of flood defences. This statement is the backbone of flood defence concepts such as the Delta dike (Knoeff and Ellen, 2011) or the “ Un-breachable” dike (Vellinga et al., 2009) which intended to increase the safety by making defences sufficiently large so that the probability of failure is close zero. To achieve it, it was proposed that the current safety standards should be reduced by a factor of 100 to cope with climate change and the increasing demographic explosion (Silva and Van Velzen, 2008). Despite the fact that this value is not necessarily correct, it does reflect the necessity of flood risk management practitioners for developing longer-term adaptation strategies which may also account for the different future uncertainties. For the case of flood defences which are specially designed for coping with sea level rise for example, the heightening of the structures is the most recurrent adaptation measure as the marginal cost per capita is significantly reduced as a function of population density (Nicholls et al., 2011). For geotechnical related uncertainties, the increase in seepage lengths and slopes became the preferred adaptation measures. Note that all mentioned strategies will require an increase in the flood defence area allocation in the future planning. This “extra space” may be more efficiently used if additional functions besides the water retaining one are included as it will not only allow to reduce the population density but it will make these solutions more financially feasible.

This is better understood from the optimal cost-benefit perspective in which the

1

Chapter

(23)

represented by the point with the lowest total costs of the cost-benefit curve. This total cost is determined as the sum of the costs of the flood defence plus the expected value of the economic damage which the flood defence avoids. For significantly larger MFFD’s, an additional term should be added (with a negative sign) to the total cost function, in order to include the added value of having the flood defence plus additional function(s) such as recreation, urban areas, commercial developments and etc. A good example of this larger structures is the “Super-Levee” (Stalenberg and Kikumori, 2008) located in Japan along the Arakawa river in the city of Tokyo (Figure 1-1). The former dike was widened and heightened for increasing its reliability. The extra generated space was urbanized so that this kind of solution becomes feasible from an economic point of view.

Figure 1-1 “Super Levee” board in Oshima Komatsugawa Park, Tokyo (Japan Times, (Brasor, 2010))

If the Dutch MFFD’s are also intended to reduce the failure probabilities by a factor of 100, a significant increase in the dimensions of the flood defence is expected (e.g. height and width). Specifically for the Netherlands, the safety standards are quite high (between 1/100,000 years and 1/300 years) with respect to other countries like for example the U.S. in which post-Katrina standards were set to flooding return periods of 1/500 years (Link, 2010). The Dutch standards are set like this as almost 60% of the country is flood prone. To achieve even stricter standards as the ones intended for the Un-breachable dike for example, massive flood defences are required which may only become economically viable by including highly profitable additional functions. Despite these facts, the present thesis is developed around the idea that MFFD’s should not be framed solely in the concept of massive financially feasible dunes, quay walls or embankments

(24)

The main idea of MFFD’s then, must be defined as “flood defences that can have additional

functions while not compromising the main water retaining one”. With this definition, a dike with

a bicycle path on top which was strengthened by sheet piling for improving its resistance is a clear and existent example of a multifunctional flood defence. Note that, the sheet pile measure can change its failure probability by a factor of 100 without increasing its original dimensions or affecting its total flooding damage cost “significantly” plus the bicycle path represents a non-water retaining function which not necessarily generates enough profit to affect the cost benefit analysis. This kind of MFFD has additional functions which are not costly or beneficial enough (in monetary terms) so that they are worth to be included either on the flood defence cost or the resultant economic damage. Nevertheless, the inclusion of this additional functions does have an important effect in their structural safety as they may interfere with the deterioration processes that occur during a flood event. These processes are commonly referred to as failure mechanisms (Vrijling, 2001). Failure mechanisms can either occur on the flood defence or its foundation. Analogously, functions are located on either the flood defence or its foundation (see Figure 1-2).

Figure 1-2 House and road in the Lekdijk in Vianen – Utrecht province

For developing the additional functions of the MFFD’s, the inclusion of hard structures inside their main “body” will be almost inevitable. Furthermore, these additional functions will require to be connected to the main infrastructure systems such as energy, sewage, communication and transport among others.

Despite the MFFD’s concept is relatively new and not yet widely implemented, the inclusion or presence of structural embedments in the flood defences is a common issue all over the world (Allsop et al., 2007; Kanning et al., 2007; Danka and Zhang, 2015; Hoffmans et al., 2015). These foreign structures are referred along this thesis as “structural embedments”.

(25)

1.2 Failure mechanisms

Failure mechanisms are modelled based on the physical phenomena that better represents the deterioration process. In the year 2007, the FLOODSITE project (Allsop et al., 2007) compiled a large inventory of case specific sets of failure mechanisms of flood defences, with the aim to improve the available knowledge. Despite the fact that all possible failure mechanisms have a chance to occur, different studies have shown that only few of them account for most of the failure registered cases (Danka and Zhang, 2015). The most relevant are overflow, wave overtopping, inner slope stability and piping erosion (Figure 1-3).

Figure 1-3 Schematics of the most dominant failure mechanisms

Overflow: Inflow of water towards the protected area due to an extreme water level

event that exceeds the height of the flood defence.

Wave Overtopping: inflow of water volumes due to wave run-up process which may

result in erosion of the crest and landward slope.

Inner Slope Stability: rotational soil mass displacement due to reduction of inner shear

strength of the soil also referred to as “Macro-stability”. It can be triggered during high water events due to the variation of the phreatic level inside the flood defence.

Piping Erosion: loss of bearing capacity of the flood defence foundation due to cavity

formation originated from soil transport due to high seepage flows.

From a historically point of view, the work of Van Baars and Van Kempen (2009) showed that, storm surges and high water levels account for 77% of the failure drivers

Overflow Wave Overtopping

(26)

and 2006. Such drivers trigger different failure mechanisms which may result in dike failure during high water events. For example, erosion of the slopes and crest due to wave overtopping corresponded to 67% of the total registered dike failures. Inner slope stability of the landward slope accounted for 5% and internal backward erosion of the foundation (also known as piping erosion) accounted for 1% of the total registered failure.

The study by Vorogushyn et al. (2009) showed that these last four failure mechanisms were also the most dominant for dike breach events in Hungary for the period between 1954 and 2004, and for the Saxony province in Germany during the flood event of August 2002. Both of these observations are partially in agreement with the ones presented by Van Baars and Van Kempen (2009) in which overtopping derived failure mechanisms (overflow, wave overtopping and wave impact) accounted for 73% of the total failures.

However, 20% in the Hungarian case and 9.5% in the German case are observed for piping erosion with respect to the 1% observed in the Dutch case. This change in percentage can be explained by the facts that flood defences have been increased in height during the last 3 decades and also some of the piping evidence was wrongly attributed to animal induced failures. A more recent study by Danka and Zhang (2015) about worldwide dike failure statistics, resulted in very similar percentages for slope stability and overtopping with respect to both Hungarian and German cases. Yet this study was based on updated data which also includes an additional 3% for failure of hard structures embedded in the flood defences and 5.8% attributed to human or animal activities among others.

While all previously mentioned studies reiteratively conclude that a large portion of the total failure probability is attributable to only four dominant failure mechanisms, it is even more remarkable how much of this portion can be attributed to erosion based failure mechanisms solely. For the study of Van Baars and Van Kempen (2009), 68% of the total failures can be attributed to piping erosion and wave overtopping. For the study of Vorogushyn et al. (2009), 89% for the Hungarian case study and 79.7% for the Saxony case study may be attributed for the same failure mechanisms. For the study of Danka and Zhang (2015), 83%. For the VNK study, the percentages depend on which dike ring system is the statistic taken into account. However, one of the general conclusions of the study is that piping failure of riverine flood defences is more important than what they estimated in the past.

(27)

1.2.1 State of the art of piping erosion

For design and assessment of piping erosion, the methods have evolved over time as more experience has been acquired from field and experimental work (Hoffmans, 2014b). Nevertheless, most of these tools consist in simplified equations with no strong physical background to support them (Bligh, 1910; Lane, 1935). In the recent years, the piping erosion state of the art has also improved significantly (Sellmeijer et al., 2011; Van Esch et al., 2013; Van Beek, 2015). More detailed description of the state of the art of these failure mechanisms is also presented in Chapters 2 and 3. Yet, this failure mechanism still lacks of through investigation for cases where structural embedments are present.

1.2.2 State of the art of overtopping

For overtopping derived failure mechanisms (either by increase of the still water level or by wave overtopping), Edelman in 1954 (Van der Meer, 2009) noticed that dikes would fail by overtopping if:

1. The crest was too low so erosion and/or flooding may occur ;

2. The quality of the materials was bad so that infiltration happened too fast resulting in slope stability;

3. The inner slope was too steep which may lead to an inner slope stability or grass cover failure;

Hence, the countermeasures taken against wave overtopping consisted in ensuring inner slopes of at least 1:3 and heightening of the dikes to a level equivalent to the wave run-up height which occurs less than 2% of the time. Due to these measures and the actual strict flood safety standards, overtopping and overflowing are nowadays the less probable failure mechanisms in the actual flood defence system according to the results of the VNK2 (VNK2, 2014). However, in the past 30 years, the overtopping derived failure mechanisms have redrawn the attention of the flood risk community due to climate change and the risk-based national policy migration (Van der Most et al., 2014). Both require a better understanding of the processes so that future measures become more cost efficient. Hence, Dutch (Van der Meer, 2002; Hoffmans et al., 2008; Van der Meer et al., 2008) and international (Pullen et al., 2007; Dean et al., 2010; Thornton et al., 2011) research initiatives were developed with the aim of increasing the knowledge in the understanding, modelling and prediction of wave overtopping effects on flood defences.

(28)

The effects derived from the inclusion of embedded structures in the wave overtopping process has become a recent research interest as significant amount of the flood defences along the Dutch and international landscapes include different transitions and structural embedments (Verheij et al., 2012; Steendam et al., 2014). The next step is to develop methods which allow to include the effects of these transitions and/or structural embedments for flood risk reduction. A first approach known as the “cumulative hydraulic load method” for wave overtopping is already in development (Hoffmans, 2015). A significant part of the development of this method is related to the grass quality of the cover.

1.2.3 State of the art of slope stability

With regard to the inner slope stability failure mechanism, greater advance has been achieved in the knowledge of this kind of failure as this deterioration process is of “paramount” importance not only for dikes but also for many other soil composed structures such as railway embankments, dams, road cuts and landfills. For the specific case of dikes, two primary adaptation measures are commonly implemented for avoiding inner slope stability failure due to high water events: the reduction of the slope and implementation of an inner berm. Note that both measures are oriented towards adding extra weight to the inner side so that the rotational mass displacement is less probable. These solutions are typically designed and tested by three main calculation methodologies: limit equilibrium, displacement based and resistance reduction methods (Weigao, 2015). The first method is the oldest of the three of them and therefore a larger amount of applications can be found in literature with respect to the other two. They are preferred by designers as their implementation is less complex with respect to Finite Element Model (FEM) based methods. For the specific case of slope stability assessments with structural embedments, the work of Paul and Kumar (1997) is an example of the implementation of limit equilibrium methods. Their results showed that failure can either occur from the localized failure of the interface between soil and structure or by a larger slip surface that includes the whole structure inside the collapsed soil mass. For the specific case of flood defences, the master research thesis of Jongerius (2016) aimed to quantify the effects on the failure probability derived from an structural house embedment. His study assessed the safety of the flood defence by assuming a limit equilibrium method conditioned to the collapsation of the embedded structure. Both studies are important stepping stones towards the fully probabilistic assessment, but the circular failure surface and the inner pressure assumptions might not be correct

(29)

1.3 Probabilistic design of flood defences

In the Netherlands, dikes were designed in the middle ages based on the most extreme water level registered in time plus one meter of freeboard in terms of height. After a disastrous flood event in 1953, a statistical approach was chosen, which allowed to extrapolate storm surge levels for flood defence design. In the later years and with the development of reliability theory in the 1980’s, the Dutch hydraulic engineers developed and implemented assessment guidelines which allowed to quantify the flooding risks taking into account the different failure mechanisms (Vrijling, 2001). From the design philosophy point of view, structures can be designed and assessed either by the allowable stress design (ASD) criteria or by the limit state design (LSD) criteria (Vrouwenvelder, 2001). The ASD ensures that the exerted stresses of the structure due to service loads do not exceed the allowable stress of the materials of which the structure is composed so that the service state is satisfied (state in which the structures still fulfils its desired function).The way to cope with the uncertainty in the material is by the use of partial safety factors (Elishakoff, 2012). The LSD uses the ultimate load and resistance criteria to ensure that both service state and limit state (state of the structure before failure of one or more of the structural components) are fulfilled. This last one acknowledges the uncertainty of both load and resistance by including partial safety factors, model uncertainty factors or even the probabilistic description of the inputs for the more advanced probabilistic methods.

1.3.1 Probabilistic design versus probabilistic

assessment

The distinction between probabilistic design and probabilistic assessment has been difficult to determine as both terms are used indiscriminately in most of the reliability related literature. One of the main reasons for this lack of term distinction is that both the design of an unconstructed structure and an existent structured may have their safety “assessed”. In other words, both the design and structure can be checked in order to determine if they comply with the legislative or recommended safety standards. Nevertheless the paper of Arangio (2012) clearly defines the distinction between the terms design and assessment by associating them to either non-existent and existent structures and the stage in which they are located inside the structural life cycle. This distinction applies to deterministic, semi-probabilistic and fully probabilistic implementation of both designs and assessments. For the present research, both terms

(30)

Probabilistic design: Determination of the required dimensions, materials and safety

factors to cope with the associated uncertainties of a future structure construction, operation and maintenance for achieving a required level of safety. The correct design choices and their inherent uncertainty representations should minimize the structural and operational costs while ensuring an associated target safety level.

Probabilistic assessment: Determination of the level of safety of an existing structure

due to its operation, maintenance and deterioration given the actual set of load and resistance conditions which are expressed in a probabilistic manner. The assessment aims to determine the actual safety level of the structure which may or not comply with its associated target safety level.

Both probabilistic design and probabilistic assessment in the present research are implemented under a limit state philosophy which ensures that the performance of a structure is evaluated until achieving it failure state.

1.3.2 Limit state design

The LSD design philosophy is widely used for the risk-based design and assessment of structures (Vrouwenvelder, 2002), as it allows to define the general limit state in a mathematical form, i.e. :

, , … Eq. 1-1

This general equation describes the failure state of a structure due to a certain deterioration process. In many situations, it is more convenient to separate it in two terms as:

, , … , , … Eq. 1-2

where represents the solicitation or “load” exerted by the structure, represents the resistance of the structure against such load and … and … represent the input

variables for estimating both and . The term represents the marginal resistance against the represented failure mechanism. When positive, the state of the structure is assumed as “not failed” whereas if negative it is assumed to be “failed”. For most of the structural designs and in particular for flood defences, the load term represents the environmental/climatological drivers that trigger the occurrence of the failure mechanism. The increase or decrease in the probability of the failure mechanisms to

(31)

uncertainty of these drivers is not possible to reduce. In consequence, the materials, the geometrical features and/or the operational guidelines become the most important tools for designers for reducing the flood risk until acceptable levels (Vrijling et al., 1998). These tools are normally reflected in the resistance term . Adaptation to climate change and/or sea level is tackled with these tools but larger uncertainty in the likelihood of the water levels is also expected in the future (Voortman and Vrijling, 2004). Consequently, both and S are uncertain which makes uncertain as well and therefore the limit state function can be expressed in terms of vectors and of random variables as:

, , … , , … Eq. 1-3

It is common to find that and may range from simple equations to highly complex computational models which may also include an stochastic component themselves (Yen, 1988). Examples of limit state equations for a significant number of different failure mechanisms of flood defences are presented in the report T04-06-01 of the FLOODsite project (Allsop et al., 2007).

1.3.3 Reliability Methods

Th e acknowledgement and complexity associated with the failure process uncertainty, represents the division line between choosing a deterministic or a probabilistic design. In particular, probabilistic design is the basis of the structural reliability discipline which aims to quantify the failure probability of structures and/or systems of structures. This is done by considering the associated uncertainties in their materials, loading conditions and operation. The “reliability” of a structure can be estimated by different methods depending on the amount of knowledge of the system; e.g. parameters stochastic nature, limit state definition and reliability function topology. In addition, there is also an obliged trade-off between accuracy and calculation time which also affects the reliability estimation (Koduru and Haukaas, 2010). With this in mind, the reliability discipline has grouped these design methods by levels (Thoft-Cristensen and Baker, 2012):

Level I: Deterministic values of and are multiplied by partial safety coefficients.

These coefficients are calibrated from fully probabilistic analysis so that level I methods represent a semi-probabilistic design basis.

Level II: Probabilistic distributions are associated to the uncertainty variables

(32)

failure surface for obtaining an approximation of the failure probability (Pf) and/or the

reliability index (β). These methods are also known as “approximated” methods.

Level III: Use of detailed numerical simulations based on the specific probabilistic

distributions associated to the input uncertain variables ( , , … and , , … ), for constructing the multi-variable joint probability density function. This type of methods allows to calculate the failure probability (Pf) and/or the reliability

index (β) from based on an “exact” failure surface.

The reliability index (β) is a common measure to express the level of reliability of a structure. Hasofer and Lind (1974) proposed to transform the function into an standardized U-space form which allowed to produce an invariant reliability β index. For these U-sapce methods, this index is be defined as the minimum distance between the origin of the bivariate ( and ) distribution and the failure surface . Such combination of and has the highest likelihood of failure.

This index is widely used in the structural reliability design and assessment as most of problems are solved by implementing level II reliability methods (approximated methods) due to their inherent computational burden. That is also one of the main reasons why reliability based design guidelines are defined in terms of target indexes in the β form. Note that all β indexes may be directly related to failure probability values estimated from a normal standardized Gaussian distribution (Φ). Hence, it is also possible to express the obtained failure probability of the level III methods in terms of β indexes as Φ (Pf).It is also possible to compare them with the target reliability β

indexes contained in the reliability design codes or legislative standards.

Level II methods are based on assumptions that the limit state function can be approximated to a linear or a second order quadratic function by means of a Taylor expansion approach (Karadeniz and Vrouwenvelder, 2003). For these methods, iterative procedures are often required in order to find the closest distance to the most probable failure point (also known as design point). The main reason for implementing these approximate methods is that they are significantly less expensive in computational terms with respect to Level III methods.

For the level III methods, the most commonly used is the Monte Carlo failure estimation in which the failure point is found by sampling the variables and/or modelling of the system according to their stochastic nature, in order to find a more exact failure probability than with level II methods. This last method in its most basic

(33)

a level III method is a better initial choice when analysing systems that may present discontinuities of the limit state function or even more important of its first derivatives which are the basis for the implementation of level II methods. It is expected that for complex structures such as MFFD’s for which the inclusion of hard structures in soft soil may result in unforeseen discontinuities of the limit state functions and therefore, only level III methods were implemented in the present thesis.

1.3.4 Fault tree analysis of flood defences

In order to estimate the total probability of failure of a structure while considering the most relevant failure mechanisms, it is common to implement a “Fault tree” method. This method allows to include the logical relationship between a series of events that may lead to the total failure of a system (Barlow, 2004). The events are linked depending on the type of relation (e.g. and, or, if) between events as seen in Figure 1-4.

Figure 1-4 Fault tree of flood defence

For flood defence systems, the system level of the fault tree is composed of structures (Vrijling, 2001). The component level is composed of representative structure sections and these sections fail due to their correspondent failure mechanisms per section. For the latest statutory assessment tools (WBI-2017, 2015) of the Dutch flood safety system, percentages of the maximum allowable failure probability per component were defined per failure mechanism based on the results of the VNK2 project (Jongejan and Calle, 2013). These failure probability budgets are referred to as “ factors” and will be

Inundation Dike failure Sea wall failure Dune failure Section 1 failure Section 2 failure Section 1 failure OR OR Wave‐overtopping failure Overflow failure Piping erosion failure Slope stability failure OR Uplift/heave failure Piping progression failure Slope stability failure Grass cover erosion OR AND OR AND Event Nomenclature Failure occurs if any sub‐event occurs Failure occurs if all sub‐events occur Continuati on of fault tree Failure event

Flood Defence Fault Tree

System Component Failure mechanisms Other mechanisms failure

(34)

Table 1-1. Failure probability budget factors per failure mechanism (WBI-2017, 2015)

factors per max. allowable Pf

Type of Flood

Defence Failure Mechanism SandyCoast Other

Dike

Wave Overtopping 0 0.24

Uplift and Piping 0 0.24

Slope stability of inner

slope 0 0.04

Revetment failure and

erosion 0 0.10 Hydraulic structures Non-Closure 0 0.04 Piping 0 0.02 Structural failure 0 0.02 Dune Erosion 0.70 0.0/0.10 Other - 0.30 0.30/0.20 TOTAL 1.0 1.0

From the suggested values for defining the maximum allowable probability it can be observed how the wave overtopping, uplift and piping and revetment failures account for 58% of the total maximum allowable failure probability. All three represent the erosion based failure mechanisms of a dike. Ideally, dikes should be designed so that the failure mechanisms probability is as low as possible and yet the coefficients are still relatively high with respect to other failure mechanisms. This may be explained by the associated uncertainty in the estimation of this failure mechanisms.

1.3.5 Dutch flood safety standards

After the major flooding event of 1953, a system of safety standards was developed and proposed by the Dutch Delta Committee. These standards (exceedance frequencies related to optimal design water levels) were obtained from the economic optimization between investment costs in flood safety and the resultant benefits of the damage reduction (Vrijling, 2001). The optimal frequency’s obtained for the primary flood defences ranged from 1/10,000 for coastal areas to higher frequencies in riverine areas between 1/1250 and 1/2000. The population growth, economic development and possible effects of climate change called for a revision of these safety standards. In the year 2006, the Ministry of Infrastructure and Environment, the Association of Regional Water Authorities and the Association of Provincial Authorities had the initiative of implementing a fully probabilistic safety assessment for all major levee systems in the Netherlands in the VNK project (Jongejan and Maaskant, 2013). One of the main

(35)

flood defence reliability were not possible to include in a probabilistic manner and that new tools should be developed for this purpose (VNK2, 2014).

More recently, a new updated cost-benefit analysis was developed in the project “WV21: water safety of the 21st_{century” (Kind, 2011). The results of this study allowed to}

redefine the required optimal safety levels of the primary flood defences. By the year 2015, a new set safety standards were available and were defined by segment instead of by dike ring as shown in Figure 1-5:

Figure 1-5 New flood standards by segment according to the Deltaprogram 2015, (Kuijken, 2015)

These standards have been formally proposed by the Deltaprogram 2015 (Kuijken, 2015), and later included approved by Dutch parliament on March 23 of 2016 as an amendment to the National Water Act. These new standards were derived taking into account factors such as the individual risk of becoming a victim of flooding (< 1/100,000 per year), the societal disruption due to flooding and the economic efficiency

1/30 1/1.000 1/3.000 1/10.000 1/30.00 1/100.000 standards [1/year]

(36)

1.4 Structural embedments

Structures like trees, buildings and cables are the most often encountered in the actual Dutch landscape and their influence on the flood defence safety has been acknowledged by the national authorities (VTV2006, 2007). In the current Dutch legislation, it is preferred to have a safe buffer zone in which additional structures should be avoided as much as possible (Zwanenburg et al., 2013). However, there are safety assessment guidelines (Beijersbergen and Spaargaren, 2009) for non-water retaining objects (NWO’s – niet waterkerende objecten in Dutch) which allow to determine if these additional structures may or not compromise the main functions of a flood defence (STOWA, 2000; VTV2006).

The method consists of elaborated flow charts in which the characteristics of the flood defence and the structural embedment are evaluated based on yes/no questions which result in a safe or unsafe definition. The method is elaborated for assessing the most typical embedments such as small buildings, trees, pipelines, cables among other random objects. An interesting characteristic of this method is that it initially evaluates the safety depending on the type of structure which is embedded. This can be interpreted as a safety philosophy approach which prioritizes the risk of the embedded structure to fail over the risk of the flood defence to fail. However, such an approach is quite understandable as the additional structures tend to be founded in the same locations of the flood defence, e.g. roads over the crest, houses inside or nearby, cables and pipes underneath or beside. Hence, the safety philosophy behind the method is to ensure that these external elements do not compromise the stability of the main elements of the flood defence such as the slopes, the crest or the core.

Different dike failure studies have also shown that the inclusion of hard structures and/or transitions between the earthen embankments and other structures originate “weak spots” in which erosion tends to start first (Kanning et al., 2007; Verheij et al., 2012; Hoffmans et al., 2015). Yet, these transitions are not explicitly included in design codes or regulations in terms of suggested design factors or safety values. By including the derived effects of the embedments in the failure mechanisms processes, it is expected to obtain better failure probability estimations of the flood defence systems while also allowing to include the embedment reliability effects in the complete safety assessments.

Flood defence failure mechanisms have specific zones of occurrence on the flood defence which allow intuiting that the location of the additional structure allow

(37)

present thesis is developed around the idea that structural embedments may be first classified by their location with respect to the flood defence so that is easier to identify which failure mechanisms they affect directly. Based on examples from the Dutch landscape, the identified influence zones were on top of the defence, inside of the body of the defence and below the flood defence (Figure 1-6). The most interesting examples of embedments by each location were identified as roads, constructions and sewer pipes respectively.

Figure 1-6 Influence zones of structural embedments

The effect of the embedment is not always related to one single failure mechanism only. For example, a house embedded inside the flood defence which may be founded deeper than the flood defence bottom will not only compromise the slope stability of the defence but will also have a significant effect on the occurrence of other failure mechanisms such as seepage through the dike, piping erosion and grass cover erosion due to wave overtopping.

Some failure mechanisms also have variable uncertainties in common derived from either the material or the geometrical characteristics which may induce correlation in their occurrence likelihood. Nevertheless, as an initial approach, it is chosen to identify cases in which one single failure mechanism is affected by one single structural embedment, or one expected correlated event affects one single failure mechanism for the sake of simplicity. For the specific cases of erosion based failure mechanisms it is acknowledged that piping erosion occurs in the foundation (“below” influence zone, Figure 1-6) whereas overtopping occurs over the flood defence (on top influence zone, Figure 1-6).

Erosion based failure mechanisms are less commonly studied with respect to deformation or retaining capacity failure mechanisms as the later ones are also found in dam engineering. This knowledge gap becomes more evident for studies related to structural embedments and erosion based failure mechanisms. Nevertheless, this information is required for the future implementation of MFFD’s.

Below Inside

(38)

1.5 Design choices and their impact on the structural

safety

During the design process, geometrical and material choices are what define the structural resistance against failure. These choices are required for both flood defence and structural embedment. However, it is expected that the design process follows the logical order of first designing the flood defence for ensuring the primary water retaining function, later include the embedment and finally assess the design. Based on this design order, the embedment of structures will change the probabilistic distribution of the and terms of the limit state equations, with respect to the initial case of the flood defence of the same dimensions and materials which has no structural embedment. The piping erosion failure mechanism is a good example of this situation as the presence of discontinuities changes the flow resistance of the aquifer and consequently the inner pressure distribution (Wang et al., 2014). Hence, the probabilistic distribution of the term will change and consequently the flood defence reliability. This is also the case of the inner slope stability safety in which the embedment of a hard structure inside the flood defence will directly affect the equilibrium state during a flood event. From this last example, it is noted that the mechanical properties of the embedded structure may contribute as well in a positive way to the flood defence stability it may add an extra resisting component against the rotational failure.

The selection and reduction of the associated variable uncertainties involved in the estimation of the load and resistance terms of a failure mechanism will also have an impact in the structure’s reliability. For example, there are failure mechanisms in which the load term statistical distribution is affected by the geometrical and geo-mechanical properties of the structure. This is the case of wave overtopping in which the overtopped water volumes erode the protecting grass cover in the landward side of a dike. If a structure is present on top of the profile, the hydrodynamics will be affected and consequently the erosion patterns will change. Or for the case of wave impact on buildings in which the dimensions of the flood defence and embedded structure may define the loading condition on the dike (Chen et al., 2015) and its probabilistic distribution (Chen et al., 2016).

For the design of MFFD’s, it is important to understand the derived effects of the structural embedment in the likelihood of the failure mechanisms. This knowledge will allow to define the degree of integration (Van Veelen et al., 2015) between structures

(39)

inclusion of embedded structures. However, their influence in the MFFD as it is out of the scope of this thesis. Hence, the remaining choices for a structural design can be classified into two groups; the material and the dimensional choices. Both design choice groups may be represented as uncertainties in the probabilistic design (Diamantidis, 1987). Based on this classification, it is defined that the most important design choices of an MFFD’s can be characterized as:

Figure 1-7 Design choices

Depending on the design choice, it is expected that a reduction of the failure probability for one or more failure mechanism may be achieved by:

1. Improving the resistance of the materials of the flood defence 2. Improving the resistance of the materials of the foundation

3. Improving the resistance of the materials of the embedded structure 4. Changing the geometry of the flood defence

5. Changing the geometry of the embedded structure 6. Changing the location of the embedded structure 7. Combinations of all of the above

The optimal selection and uncertainty assumption of each of the choices can be used for accurately estimating and improving the MFFD safety. For riverine dikes, piping erosion and wave overtopping are influenced by all 7 design choices.

?

Defence Materials Foundation Materials Embedment Materials

?

Defence Geometry

?

Embedment Size Embedment Location

?

? Material choices Dimensional choices

(40)

1.5.1 Model selection for probabilistic design and

assessment

The selection of the model that represents the failure situation better represents another important design choice. Nowadays, numerical models are capable of representing reality in different levels of complexity. Models are representations of physical processes in which assumptions allow to simplify (or increase) their computational complexity (Brooks and Tobias, 1996). Model complexity depends on the number of processes involved, their time dependency, the interaction between processes and the number of spatial dimensions in which the processes are modelled, among others. According to the study of Wainwright and Mulligan (2005), models may be classified based on their complexity as empirical, conceptual and physically based. For the four main flood defence failure mechanisms, reference studies in which different modelling complexities are implemented are summarized in Table 1-2. After the fast advance of computational intelligence, a new kind of modelling approach emerged which is formerly known as “data driven”. In this type of modelling, the processes are recreated based only on algorithms built based on the input and output collected data of the physical process, disregarding the physical representation of the process itself (Solomatine et al., 2009).

Table 1-2. Failure mechanisms modelling studies with different complexities Failure mechanism Empirical models Conceptual models Physically based models Data Driven models

- (Hewlett et al., ₁₉₈₇₎ (Pontillo et al., 2010) (Aguilar-López _{et al., 2014)}

(Van der Meer, 2002) (Schüttrumpf and Oumeraci, 2005) (Quang and Oumeraci, 2012) and (Aguilar-López et al Chapter 4) (van Gent et al., 2007) (Lane, 1935) and (Sellmeijer et al., 2011) (Sellmeijer, 1988) Wang et al. (2014); (Aguilar-López et al., 2016a) (Kaunda, 2015; Aguilar-López et al., 2016a) (Johnson et

al., 1999) (Van et al., 2005) (Moellmann et al., 2011) (Kingston, 2011)

Overflow

Wave Overtopping

Piping Erosion

Probabilistic safety assessment of multi-functional flood defences

Juan Pablo Aguilar López

Probabilistic Safety Assessment of

Multi-functional Flood Defences

Probabilistic Safety Assessment of

Multi-functional Flood Defences

PROBABILISTIC SAFETY ASSESSMENT

OF MULTI-FUNCTIONAL FLOOD DEFENCES

DISSERTATION

Juan Pablo Aguilar López

‘By three methods we may learn wisdom: First, by reflection, which is the

noblest; Second, by imitation, which is the easiest; and third by experience,

which is the bitterest.’

Confucius (551 BC - 479 BC)

Preface

Contents

Summary

Samenvatting

1.

Introduction

1.1 Multi-functional flood defences (MFFD’s)

1

Chapter

1.2 Failure mechanisms

1.2.1 State of the art of piping erosion

1.2.2 State of the art of overtopping

1.2.3 State of the art of slope stability

1.3 Probabilistic design of flood defences

1.3.1 Probabilistic design versus probabilistic

assessment

1.3.2 Limit state design

1.3.3 Reliability Methods

1.3.4 Fault tree analysis of flood defences

Flood Defence Fault Tree

1.3.5 Dutch flood safety standards

1.4 Structural embedments

1.5 Design choices and their impact on the structural

safety

?

?

?

?

?

1.5.1 Model selection for probabilistic design and

assessment